diff options
author | Ole Troan <ot@cisco.com> | 2022-03-17 11:58:38 +0100 |
---|---|---|
committer | Ole Tr�an <otroan@employees.org> | 2022-04-21 10:35:58 +0000 |
commit | 5297447bd64ab253ab3ab3e144605dd39f995f12 (patch) | |
tree | d351435893287349ebb9ca5736a627ff53cc28f9 /src/plugins/nat | |
parent | e0c875551fa0cd49131671be0f521801e06764f8 (diff) |
nat: tweak rfc7857 tcp connection tracking
The RFC7857 state machine introduced in 56c492a is a trade-off.
It tries to retain sessions as much as possible and also offers
some protection against spurious RST by re-establishing sessions if data
is received after the RST. From experience in the wild, this algorithm is
a little too liberal, as it leaves too many spurious established sessions
in the session table.
E.g. a oberserved pattern is:
client server
<- FIN, ACK
ACK ->
ACK ->
RST, ACK ->
With the current state machine this would leave the session in established state.
These proposed changes do:
- require 3-way handshake to establish session.
(current requires only to see SYNs from both sides)
- RST will move session to transitory without recovery if data is sent after
- Only a single FIN is needed to move to transitory
Fixes: 56c492aa0502751de2dd9d890096a82c5f04776d
Type: fix
Signed-off-by: Ole Troan <ot@cisco.com>
Change-Id: I92e593e00b2efe48d04997642d85bd59e0eaa2ea
Signed-off-by: Ole Troan <ot@cisco.com>
Diffstat (limited to 'src/plugins/nat')
-rw-r--r-- | src/plugins/nat/nat44-ed/nat44_ed.c | 2 | ||||
-rw-r--r-- | src/plugins/nat/nat44-ed/nat44_ed.h | 18 | ||||
-rw-r--r-- | src/plugins/nat/nat44-ed/nat44_ed_format.c | 29 | ||||
-rw-r--r-- | src/plugins/nat/nat44-ed/nat44_ed_inlines.h | 304 |
4 files changed, 61 insertions, 292 deletions
diff --git a/src/plugins/nat/nat44-ed/nat44_ed.c b/src/plugins/nat/nat44-ed/nat44_ed.c index f2d2c715760..9c797536fb6 100644 --- a/src/plugins/nat/nat44-ed/nat44_ed.c +++ b/src/plugins/nat/nat44-ed/nat44_ed.c @@ -2456,8 +2456,6 @@ nat44_plugin_enable (nat44_config_t c) nat44_ed_db_init (sm->max_translations_per_thread, sm->translation_buckets); - nat44_ed_init_tcp_state_stable (sm); - nat_affinity_enable (); nat_reset_timeouts (&sm->timeouts); diff --git a/src/plugins/nat/nat44-ed/nat44_ed.h b/src/plugins/nat/nat44-ed/nat44_ed.h index 05503a475c2..5b5b2ec8cfd 100644 --- a/src/plugins/nat/nat44-ed/nat44_ed.h +++ b/src/plugins/nat/nat44-ed/nat44_ed.h @@ -123,14 +123,10 @@ typedef enum typedef enum { - NAT44_ED_TCP_FLAG_NONE = 0, - NAT44_ED_TCP_FLAG_FIN, + NAT44_ED_TCP_FLAG_FIN = 0, NAT44_ED_TCP_FLAG_SYN, - NAT44_ED_TCP_FLAG_SYNFIN, NAT44_ED_TCP_FLAG_RST, - NAT44_ED_TCP_FLAG_FINRST, - NAT44_ED_TCP_FLAG_SYNRST, - NAT44_ED_TCP_FLAG_SYNFINRST, + NAT44_ED_TCP_FLAG_ACK, NAT44_ED_TCP_N_FLAG, } nat44_ed_tcp_flag_e; @@ -145,15 +141,8 @@ typedef enum typedef enum { NAT44_ED_TCP_STATE_CLOSED = 0, - NAT44_ED_TCP_STATE_SYN_I2O, - NAT44_ED_TCP_STATE_SYN_O2I, NAT44_ED_TCP_STATE_ESTABLISHED, - NAT44_ED_TCP_STATE_FIN_I2O, - NAT44_ED_TCP_STATE_FIN_O2I, - NAT44_ED_TCP_STATE_RST_TRANS, - NAT44_ED_TCP_STATE_FIN_TRANS, - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O, - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I, + NAT44_ED_TCP_STATE_CLOSING, NAT44_ED_TCP_N_STATE, } nat44_ed_tcp_state_e; @@ -336,6 +325,7 @@ typedef CLIB_PACKED(struct u16 ext_host_nat_port; /* TCP session state */ + u8 tcp_flags[NAT44_ED_N_DIR]; nat44_ed_tcp_state_e tcp_state; /* per vrf sessions index */ diff --git a/src/plugins/nat/nat44-ed/nat44_ed_format.c b/src/plugins/nat/nat44-ed/nat44_ed_format.c index f643d1c6056..ee3e925e529 100644 --- a/src/plugins/nat/nat44-ed/nat44_ed_format.c +++ b/src/plugins/nat/nat44-ed/nat44_ed_format.c @@ -303,34 +303,11 @@ format_nat44_ed_tcp_state (u8 *s, va_list *args) case NAT44_ED_TCP_STATE_CLOSED: s = format (s, "closed"); break; - case NAT44_ED_TCP_STATE_SYN_I2O: - s = format (s, "SYN seen in in2out direction"); - break; - case NAT44_ED_TCP_STATE_SYN_O2I: - s = format (s, "SYN seen in out2in direction"); - break; case NAT44_ED_TCP_STATE_ESTABLISHED: - s = format (s, "SYN seen in both directions/established"); - break; - case NAT44_ED_TCP_STATE_FIN_I2O: - s = format (s, "FIN seen in in2out direction"); - break; - case NAT44_ED_TCP_STATE_FIN_O2I: - s = format (s, "FIN seen in out2in direction"); - break; - case NAT44_ED_TCP_STATE_RST_TRANS: - s = format (s, "RST seen/transitory timeout"); - break; - case NAT44_ED_TCP_STATE_FIN_TRANS: - s = format (s, "FIN seen in both directions/transitory timeout"); - break; - case NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I: - s = format (s, "FIN seen in both directions/transitory timeout/session " - "reopening in out2in direction"); + s = format (s, "established"); break; - case NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O: - s = format (s, "FIN seen in both directions/transitory timeout/session " - "reopening in in2out direction"); + case NAT44_ED_TCP_STATE_CLOSING: + s = format (s, "closing"); break; case NAT44_ED_TCP_N_STATE: s = format (s, "BUG! unexpected N_STATE! BUG!"); diff --git a/src/plugins/nat/nat44-ed/nat44_ed_inlines.h b/src/plugins/nat/nat44-ed/nat44_ed_inlines.h index f50ecf7ffcc..b99d152da26 100644 --- a/src/plugins/nat/nat44-ed/nat44_ed_inlines.h +++ b/src/plugins/nat/nat44-ed/nat44_ed_inlines.h @@ -176,21 +176,7 @@ nat_get_icmp_session_lookup_values (vlib_buffer_t *b, ip4_header_t *ip0, always_inline int nat44_ed_tcp_is_established (nat44_ed_tcp_state_e state) { - static int lookup[] = { - [NAT44_ED_TCP_STATE_CLOSED] = 0, - [NAT44_ED_TCP_STATE_SYN_I2O] = 0, - [NAT44_ED_TCP_STATE_SYN_O2I] = 0, - [NAT44_ED_TCP_STATE_ESTABLISHED] = 1, - [NAT44_ED_TCP_STATE_FIN_I2O] = 1, - [NAT44_ED_TCP_STATE_FIN_O2I] = 1, - [NAT44_ED_TCP_STATE_RST_TRANS] = 0, - [NAT44_ED_TCP_STATE_FIN_TRANS] = 0, - [NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O] = 0, - [NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I] = 0, - [NAT44_ED_TCP_N_STATE] = 0, - }; - ASSERT (state <= ARRAY_LEN (lookup)); - return lookup[state]; + return state == NAT44_ED_TCP_STATE_ESTABLISHED ? 1 : 0; } always_inline u32 @@ -745,251 +731,74 @@ nat44_ed_session_reopen (u32 thread_index, snat_session_t *s) s->total_bytes = 0; } -always_inline void -nat44_ed_init_tcp_state_stable (snat_main_t *sm) -{ - /* first make sure whole table is initialised in a way where state - * is not changed, then define special cases */ - nat44_ed_tcp_state_e s; - for (s = 0; s < NAT44_ED_TCP_N_STATE; ++s) - { - int i; - for (i = 0; i < NAT44_ED_N_DIR; ++i) - { - int j = 0; - for (j = 0; j < NAT44_ED_TCP_N_FLAG; ++j) - { - sm->tcp_state_change_table[s][i][j] = s; - } - } - } - - /* CLOSED and any kind of SYN -> HALF-OPEN */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_SYN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_SYN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_CLOSED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_SYN_O2I; - - /* HALF-OPEN and any kind of SYN in right direction -> ESTABLISHED */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_SYN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; - - /* ESTABLISHED and any kind of RST -> RST_TRANS */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_RST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_RST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_FINRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_FINRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_RST_TRANS; - - /* ESTABLISHED and any kind of FIN without RST -> HALF-CLOSED */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_FIN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_FIN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_ESTABLISHED][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_O2I; - - /* HALF-CLOSED and any kind of FIN -> FIN_TRANS */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_FINRST] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_FINRST] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_I2O][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_FIN_TRANS; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_O2I][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_FIN_TRANS; - - /* RST_TRANS and anything non-RST -> ESTABLISHED */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_NONE] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_NONE] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_FIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_RST_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - - /* FIN_TRANS and any kind of SYN -> HALF-REOPEN */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_I2O] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_TRANS][NAT44_ED_DIR_O2I] - [NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I; - - /* HALF-REOPEN and any kind of SYN in right direction -> ESTABLISHED */ - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O] - [NAT44_ED_DIR_O2I][NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I] - [NAT44_ED_DIR_I2O][NAT44_ED_TCP_FLAG_SYN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O] - [NAT44_ED_DIR_O2I][NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I] - [NAT44_ED_DIR_I2O][NAT44_ED_TCP_FLAG_SYNRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O] - [NAT44_ED_DIR_O2I][NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I] - [NAT44_ED_DIR_I2O][NAT44_ED_TCP_FLAG_SYNFIN] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O] - [NAT44_ED_DIR_O2I][NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; - sm->tcp_state_change_table[NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I] - [NAT44_ED_DIR_I2O][NAT44_ED_TCP_FLAG_SYNFINRST] = - NAT44_ED_TCP_STATE_ESTABLISHED; -} - -/* TCP state tracking according to RFC 7857 (and RFC 6146, which is referenced - * by RFC 7857). Our implementation also goes beyond by supporting creation of - * a new session while old session is in transitory timeout after seeing FIN - * packets from both sides. */ +/* + * "Some rise by SYN, and some by virtue FIN" - William Shakespeare + * TCP state tracking patterned after RFC 7857 (and RFC 6146, which is + * referenced by RFC 7857). In contrast to the state machine in RFC7857 we only + * transition to ESTABLISHED state after seeing a full 3-way handshake (SYNs + * and ACKs in both directions). RFC7857 as a means of protecting against + * spurious RSTs closing a session, goes back to ESTABLISHED if a data packet + * is received after the RST. This state machine will leave the state in + * transitory if RST is seen. Our implementation also goes beyond by supporting + * creation of a new session while old session is in transitory timeout after + * seeing FIN packets from both sides. + */ always_inline void nat44_set_tcp_session_state (snat_main_t *sm, f64 now, snat_session_t *ses, u8 tcp_flags, u32 thread_index, nat44_ed_dir_e dir) { snat_main_per_thread_data_t *tsm = &sm->per_thread_data[thread_index]; - nat44_ed_tcp_flag_e flags = - tcp_flags & (TCP_FLAG_SYN | TCP_FLAG_FIN | TCP_FLAG_RST); + u8 old_flags = ses->tcp_flags[dir]; + ses->tcp_flags[dir] |= + tcp_flags & (TCP_FLAG_FIN | TCP_FLAG_SYN | TCP_FLAG_RST | TCP_FLAG_ACK); + if (old_flags == ses->tcp_flags[dir]) + return; u8 old_state = ses->tcp_state; - ses->tcp_state = sm->tcp_state_change_table[ses->tcp_state][dir][flags]; - if (old_state != ses->tcp_state) + switch (old_state) { - if (nat44_ed_tcp_is_established (ses->tcp_state)) + case NAT44_ED_TCP_STATE_CLOSED: + // ESTABLISHED when a SYN and ACK is seen from both sides + if ((ses->tcp_flags[NAT44_ED_DIR_I2O] & + ses->tcp_flags[NAT44_ED_DIR_O2I]) == (TCP_FLAG_SYN | TCP_FLAG_ACK)) { - if (NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O == old_state || - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I == old_state) - { - nat44_ed_session_reopen (thread_index, ses); - } + ses->tcp_state = NAT44_ED_TCP_STATE_ESTABLISHED; ses->lru_head_index = tsm->tcp_estab_lru_head_index; } - else + break; + case NAT44_ED_TCP_STATE_ESTABLISHED: + // CLOSING when a FIN is seen from either side or session has been RST + if ((ses->tcp_flags[dir] & TCP_FLAG_FIN) || + (ses->tcp_flags[dir] & TCP_FLAG_RST)) { - if (NAT44_ED_TCP_STATE_ESTABLISHED == old_state) - { // need to update last heard otherwise session might get - // immediately timed out if it has been idle longer than - // transitory timeout - ses->last_heard = now; - } + ses->tcp_state = NAT44_ED_TCP_STATE_CLOSING; + ses->tcp_flags[NAT44_ED_DIR_I2O] = 0; + ses->tcp_flags[NAT44_ED_DIR_O2I] = 0; + // need to update last heard otherwise session might get + // immediately timed out if it has been idle longer than + // transitory timeout + ses->last_heard = now; ses->lru_head_index = tsm->tcp_trans_lru_head_index; } - ses->last_lru_update = now; - clib_dlist_remove (tsm->lru_pool, ses->lru_index); - clib_dlist_addtail (tsm->lru_pool, ses->lru_head_index, ses->lru_index); + break; + case NAT44_ED_TCP_STATE_CLOSING: + // Allow a transitory session to reopen + if ((ses->tcp_flags[NAT44_ED_DIR_I2O] & + ses->tcp_flags[NAT44_ED_DIR_O2I]) == (TCP_FLAG_SYN | TCP_FLAG_ACK)) + { + nat44_ed_session_reopen (thread_index, ses); + ses->tcp_state = NAT44_ED_TCP_STATE_ESTABLISHED; + ses->lru_head_index = tsm->tcp_estab_lru_head_index; + } + break; } + if (old_state == ses->tcp_state) + return; + ses->last_lru_update = now; + clib_dlist_remove (tsm->lru_pool, ses->lru_index); + clib_dlist_addtail (tsm->lru_pool, ses->lru_head_index, ses->lru_index); } always_inline void @@ -1012,13 +821,8 @@ always_inline void nat44_session_update_counters (snat_session_t *s, f64 now, uword bytes, u32 thread_index) { - if (NAT44_ED_TCP_STATE_RST_TRANS != s->tcp_state && - NAT44_ED_TCP_STATE_FIN_TRANS != s->tcp_state && - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_I2O != s->tcp_state && - NAT44_ED_TCP_STATE_FIN_REOPEN_SYN_O2I != s->tcp_state) - { - s->last_heard = now; - } + // regardless of TCP state, reset the timer if data packet is seen. + s->last_heard = now; s->total_pkts++; s->total_bytes += bytes; } |